This action might not be possible to undo. Are you sure you want to continue?
Vibration Institute Annual Training Seminar
Advanced Field Balancing Techniques
Ray D. Kelm, P.E. Kelm Engineering Danbury, TX firstname.lastname@example.org
Vibration in rotating machinery is commonly the result of mechanical faults including mass unbalance, coupling misalignment, loose components, and many other causes. Improving the levels of vibration should always include elimination of the source of vibration and not addressing the symptom by making balance corrections. Mass unbalance will produce vibration due to the force generated by the eccentric weight. This force will be imposed at the running speed of the shaft, and depends on the amount of eccentric mass , the eccentricity of the weight , and the frequency of rotation . In more common terms the unbalance is defined by the eccentric weight, mounting radius, and shaft speed. The observed vibration signature will show elevated amplitudes at 1xRPM and no other significant frequencies when rotor unbalanced is the main fault.
Unfortunately, other common faults can also generate high levels of vibration at 1xRPM including coupling misalignment, looseness, rotor bows, and a variety of other sources. In some cases, these faults will produce other symptoms that can suggest corrections other than balancing should be done. Yet in many cases, balancing may be the chosen course of action for lowering vibration amplitudes even though it is not the source of vibration. Once it is determined that balance corrections should be made, the balancing process includes measuring reference vibration, adding trial weights, observing the response due to trial weights, and using the response characteristics to determine the location of balance correction weights to reduce vibration to an acceptable level. Once a field balance has been completed on a machine (or similar machine) the response data from adding trial weights can be used to calculate future balance corrections using a one shot method (no trial weights required). This paper provides a description of the process used for field balancing of a variety of rotors including single and multi-plane methods. Details include a theoretical description along with some practical suggestions, pitfalls, and examples for each of the balance methods.
Determining the Need for Field Balancing
Prior to attempting any balance corrections, a proper vibration analysis should be done to determine the likelihood that the machine is in fact out of balance. Making balance corrections to a machine with some other fault can in many cases reduce the vibration amplitudes. However, if balance corrections are made to a machine that is not out of balance to start with, the forces generated by the fault will still exist even though balance corrections may reduce the amplitude at some measurement points. Unbalance will generate forces and corresponding vibration response at 1xRPM. For example, if the rotor speed is 3600 RPM, unbalance will produce vibration at 3600 CPM (60 Hz). The FFT plot in figure 1 shows an example of an unbalanced rotor, with vibration shown at the shaft speed and minimal vibration content elsewhere. Since this data was recorded from installed proximity probes and some minor imperfections exist in the shaft surface, there is a small 2xRPM component due to runout.
Figure 1 - Typical FFT from Unbalance
Another common characteristic of unbalance is the vibration phase difference that should be about 90° between vertical and horizontal locations, with the phase angles increasing with rotating as shown on the orbit plot in figure 2. This display shows that the phase difference between two orthogonal probes (90° apart) is nearly 90°. The actual phase difference is 70° as shown in the notes on the plot which can occur due to a variety of reasons. Most common machinery with horizontal rotors will also show higher amplitude in the horizontal direction compared to vertical. Fans and motors will generally have a vertical/horizontal difference between 1:1 and 1:10 (horizontal possibly 10 times higher than vertical), which is usually dependant on the flexibility of the bearing housing and/or pedestals. This indicates that most equipment will be stiffer vertically than horizontal, which makes sense since vertical motion will be
can be corrected using a single plane balance method. indicated by the same phase angle in the same measurement direction at both bearings. If an orbit plot is viewed using seismic probes (accelerometers) on a bearing housing.Orbit for Unbalance . The phase difference from one bearing to the other (inboard to outboard) can be used to assess if the unbalance is static or dynamic. multiple balance correction planes are often required. In many cases static unbalance. it is not unusual to see a large horizontal movement and little vertical movement as shown in figure 3.resisted by tension/compression of the support where horizontal results in bending of the support. Figure 2 . When phase angles differ from one bearing to the other.
Orbit for Casing Sensors for Horizontally Flexible Machine .Figure 3 .
• “Field” Balancing – Most rotating components are balanced during the manufacturing process to include balance corrections of individual components (hubs. Vibration Vector – Vibration amplitudes for the balancing is documented as amplitude and phase. Corrections made in a manufacturer’s shop will normally be done using a balance machine where the part is either mounted on a shop mandrel or else the entire rotor is balanced. Multi-plane Balance – This includes a balance process where it is necessary to make balance corrections at 2 or more axial locations along the shaft where each plane will have normally different weights and phase angles. The corresponding phase is the phase angle lag for 1xRPM vibration with the lag defined as the lag angle between the firing (trigger) of the tachometer reference and the peak of the 1xRPM vibration amplitude. Effect – This is a term used to describe the ratio between the amount of weight added and the vector response of the vibration and can be the same as the balance coefficient. it would be prudent to discuss some common terms related to balancing. Single Plane Balance – This includes a balance process where corrections are made in a single axial location or at multiple locations using equal amplitude weights at the same phase (also called a “static” correction) that can successfully reduce vibration to acceptable levels. are common. etc.the response vector angle.Definition of Terms Prior to getting into the balance details. There appears to be no consistent definition for the effect. Any vibration amplitude unit can be used although considerations for phase difference between acceleration. Balance Coefficient – The balance sensitivity for a given sensor to a known balance correction weight. etc. impellers. and adding field corrections weights to improve the machine vibration at bearing housings or other locations. The use of the term “effect” is generally limited to single plane balancing. The amplitude is the magnitude of the 1xRPM vibration that is filtered using some electronic or software method. gram/(in/sec). Vibration vectors are most commonly displayed using a polar graph to allow simultaneous display of both magnitude and phase. but units of Lb/mil. the weight is the amount of added or removed weight (weight change) and the response is the vector change (tip to tip on a polar plot). ounces/mil. Multi-plane balancing is generally described using influence coefficients instead of effects. Units are weight/response with units of grams/mil. and includes a phase angle that is the trial weight angle location + 180° . In determining the effect. Units of Weight/Response or Response/Weight are used. field balancing involves using vibration measurements on fully assembled machines that are usually in their final service location. mils/Lb. velocity and displacement should be considered when locating trial weights.) and to make corrections to assemblies of parts (rotors). gram/mil. • • • • • . In contrast. etc.
Balance Lag Angle – This term is used in balancing to define the lag or delay between the passing of an unbalance weight (heavy spot) past a sensor and the peak of the vibration in displacement units (high spot) indicated from the same sensor location. Phase angles change between acceleration. and the phase angle is usually documented as the phase angle for the weight on one end of the rotor (normally the drive end). Equipment with proximity probes may demonstrate some apparent lag if there is high housing vibration since the probes measure bearing housing to shaft vibration. No standard exists for documenting the angle (which weight to use) or for the amount of weight (1 or two times the actual correction weight). Some additional phase lag is likely on bearing housing seismic measurements due to damping between the shaft and the housing particularly with oil film bearings. or the phase lag angle between a stationary point on the shaft (leading edge of a piece of reflective tape or of a notch in the shaft) and the peak of the vibration signal. The amount of weight would either be referred to as the total weight (2 times a single weight) or the weight of a single weight. These terms are normally calculated using the multi-plane balance method described in this paper. Electrical lag (instrument lag) is not a realistic concern with modern data acquisition equipment including laser or optical tachometers and accelerometers or proximity probes. It is reported in degrees. and 110-180° for rotors operating above a critical speed. This can be estimated using the balance coefficients or influence coefficients and would be equal to a calculated “trim” correction weight determined from the final vibration levels after balancing is complete. This phase angle difference must be accounted for when using the phase angle to determine the locations of trial weights. velocity and displacement by 90° each. No standard method is currently defined. This weight mounting method is commonly used for rotors that are reasonably symmetric and that may have similar vibration on each bearing that is in phase. Some instrument lag will exist with velocity coil type sensors. Residual Unbalance – This is the amount of unbalance that remains in the rotating assembly after adequate balance corrections have been made. Phase Angle – This paper only considers what is referred to as stationary phase. Influence coefficients are best considered in a matrix form with rows and columns corresponding to sensor locations and balance correction planes. with common values of 10-70° for rotors that operate below a critical speed. the results are identical to the balance coefficients for each sensor that can be obtained using a graphical method. Couple Balance Correction – Also referred to as a dynamic balance correction. This is similar to the static correction except that the identical weight is located 180° apart from end to end. The amount of weight added may be referred to as the total weight (2 times the weight of a single weight) or as the amount of a single weight. When calculated under conditions where only a single balance correction plane is used. Static Balance Correction – This is a description of a balance correction made to a rotor using two balance correction weights of equal amplitude that are located on opposite ends of a rotor or impeller.• • • • • • Influence Coefficients – The ratio of vector change in vibration for a given sensor and the amount of weight added to a specific balance correction plane. .
particularly ISO 1940. These filters can produce amplitude and phase errors with some instruments particularly below about 10 Hz (600 RPM). Trial weights can be risky to add particularly for rotors that are highly sensitive to correction weight. it would be more logical to balance using velocity measurements even though the single frequency vibration can be easily converted between unit types. Obviously if the balance specification requires verification of a velocity reading. Trim Balance – A balance process often called a “one shot” when previous balance data (balance coefficients) are used to calculate a shot without adding trial weights. Projected Vibration – Calculated vibration response that is expected after a shot is added to the rotor. Eccentricity – The radial distance of the unbalance or the balance correction weight relative to the rotor center of rotation (radius to the correction weight). One particular case where using displacement may not be desired is when the test instrumentation applies some integration filters to prevent integration noise known as a “ski slope” near the left side of the FFT plot. Added weights are normally documented as positive. use rotor eccentricity to determine the allowable residual unbalance. Some standards. There may be some cases where a unit other than displacement would be preferred. a rotor eccentricity (instead of balance weight eccentricity) is described which is determined by the level of residual unbalance divided by the rotor weight. Units of Measure Field balancing can be accomplished using any vibration measurement unit that is proportional to the unbalance force. and removed weight as negative weight amplitudes at the removed angle. . Shot or Balance Shot – A term used to describe the process of adding trial final correction weights followed by operation to determine the vibration response. Rotor eccentricity is the amount of radial offset that the entire rotor would have to be eccentric by to produce the same unbalance force that would be generated by a specific correction weight mounted at a specific radius. For many users. Estimation methods have been presented in literature suggesting trial weight amplitudes that will produce dynamic forces that is less than or equal to 10% of the rotor static weight. This method is also used after a final correction weight is added to determine a smaller trim weight that can be added without disturbing the final weights that is expected to further reduce the vibration amplitudes.• • • • • • Correction Weight – The amount of weight added to or removed from a rotor at a specific angular location. Common units of measure will include displacement or velocity although there is no technical reason why acceleration could not be used as well. Trial Weight – This is a temporary correction weight added (or removed) for the purpose of characterizing the response of the rotor. displacement measurements will result in more logical positioning of trial weights since the displacement phase always identifies the “high spot”. A removal of weight at one location is identical in function to adding the same amount of weight to the opposite side of the rotor. In some cases.
This is demonstrated on the polar plot in figure 4 as well as time waveform plots for the same data in different units in figure 5.Figure 4 . Any time a signal is integrated. the phase shifts by 90°. velocity and displacement. The phase .Polar Plot Using Different Units of Measure for Identical Data One very significant point to consider is the phase difference between acceleration.
In general. but these can produce significant problems during a particular field balance. accurate/repeatable test measurements. In this scenario. Linear response of the system simply states that if I get a response of 1 mil for a certain weight size that I should get a 2 mil response for twice that amount of added weight. This nonlinear reaction is the primary cause for overshooting (driving the response 180° past the center of a polar plot) when a final correction weight is added if the reference vibration was high (>>0.Time Waveform for Different Units of Measure Balancing Assumptions There are some basic assumptions that are made when doing field balancing. and consistent weight placement. This will normally be mostly true unless the rotor is severely out of balance. These sound like simple assumptions.direction change will show that acceleration leads velocity by 90°. the trial weight will characterize the response in the nonlinear range producing a calculated weight that is excessive or that is moderately off in phase angle. the phase measurement is normally converted to a displacement phase (the actual “high spot”) with weights located relative to that phase with some consideration for the lag angle between the weight (“heavy spot”) and the peak phase (“high spot”). velocity leads displacement by 90°.3 in/sec). When using the vibration phase to determine and angular position of trial weights. The test instrument used should properly display the phase angles to be relative to the unit of vibration amplitude so that the phase lag shown is the lag angle between the tachometer pulse and the peak of the filtered vibration waveform when displayed in the unit of measure. In some . Figure 5 . and due to double integration acceleration leads displacement by 180°. These include: linear response.3 in/sec peak will begin to produce non-linearity so that it will take more (or less) weight to produce the same about of vibration change. vibration in excess of about 0.
heat levels. Variable readings can be even more difficult to identify. ANSI S2. the rotor should be clearly marked with angular positions so that all additional weight additions are properly made relative to the assumed phase angle for the initial trial weights. loading differences. it is absolutely mandatory to assure that the same machine conditions are used during vibration measurement for each balance run including rotor speed.1 in/sec pk on bearing housings Antifriction bearing motors – < 0. To eliminate this risk. etc.cases. etc. you will not get dependable balance calculations. and can be caused by thermally induced rotor bows. turbines.19. When field balancing equipment. Being patient and acquiring consistent data will generally produce quicker results on most balance jobs than rushing to get finished at the cost of less accurate data. the accuracy of the readings can be affected by the placement or movement of the vibration sensors or movement of the tachometer (phase) reference sensor.05 in/sec pk on bearing housings Antifriction bearing centrifugal fans/blowers – < 0. Accurate and repeatable vibration amplitudes sound obvious. or the tachometer is pointing to the reflective tape but 10° away from the previous run. alignment offsets due to variations in casing temperatures.08 in/sec peak on the bearing housing or motor frame . Balance Tolerances Field balancing tolerances are not well defined in available literature except in some requirements defined by end users or equipment manufacturers. these specifications generally do not specifically define a level of vibration or residual unbalance in an assembled machine at operating speed. once a balance process is started. However. So if somebody moves a sensor between runs. the following tolerances are suggested as reasonable targets for some machine types using 1xRPM filtered vibration amplitudes: • • • • Turbomachinery (API compressors. the possible phase error between the rotor weights and the actual tachometer position is not significant since the balance corrections are all made relative to the trial weights. Therefore. Placement of trial weights and final correction weights can be inaccurate if a care is not used in weight placement. If the weight additions are done in this fashion (particularly if there is not previous balance data and you aren’t trying to reuse the balance response in the future). machine load.) – Mils Pk-Pk < using shaft to bearing housing relative vibration Large sleeve bearing equipment without proximity probes – < 0. Currently available test instruments will produce consistent amplitude and phase readings for the same input signals in most cases. Determining the actual location of the tachometer firing (normally the leading edge of the reflective tape) relative to a position on the rotor where the correction weights are added can produce large phase errors. etc. process temperatures. and API. the phase of the balance coefficients can change with vibration amplitude making weight placement errors as high as 20-40°. However. Many specifications are available that define allowable residual unbalance levels for a shop balance such as ISO 1940.
However. previous balance weights may have come off due to being poorly installed or improper materials (corrosion). The final correction weights will include: • • • • • Welded plates Engineered weights (same as above) Clamp on weights (when welding is not practical) Removing weight (drilling holes. washers. field balancing is ideally done until the phase readings become unstable due to the low amplitudes of vibration. practical field balancing is frequently finished based on a limited number of balance shots. etc. Some trial weights can include: • • • • Clamp on balance weights (these are commercially available in a wide variety of weights.) on machines that have removable balance weights such as power turbines and generators Final correction weights are by definition intended to be left permanently on the rotor to correct for unbalance. it is generally desirable to have some method of making trial weight moves that can be easily mounted and removed once final weights are determined. The other benefit of this type of visual inspection is the benefit of looking at previous balance weights that have been used as a mental reference point for the selection of trial weight magnitudes.In general. shapes. It should be noted that for cases where very high levels of vibration are observed in the initial reference run. Particularly on fans. cracks or damage to impellers should normally be repaired prior to making balance corrections. . etc. nuts. by time or on the capability and insistence of the balancer. Please review the trial weight selection section for additional guidance in selecting trial weights. etc. In some cases. the nonlinear response may require starting the balance process over once the vibration is at reasonable levels. In addition. Weight Corrections Balance weight corrections can be done in a number of ways.) I would always prefer to inspect the rotor for causes of unbalance to determine if there is a good reason for the rotor being out of balance and to help identify other problems that could exist. grinding. installation method and material type) Balancing putty (the same stuff used on a shop balance machine) Added bolts/washers/nuts Engineered weights (dovetail slots. the presence of a lot of fouling may warrant cleaning it off instead of field balancing.) Bolted on weights (added bolts. balance plugs. When performing a field balance.
Numerous published works describe how to perform single plane balancing using graphical methods. Single plane balancing can be very useful even for rotors that would normally be corrected using a two plane balance method when vibration amplitudes and phase readings allow static or couple corrections alone. It is also common after addition of the final weight (CW) to need to add a smaller trim weight that is added to further reduce the vibration while leaving the final weight in place. Based on the huge benefit of an intuitive approach to single and multi-plane field balancing. a computer program can also be used for single plane balancing. The basic balance process requires measurement of a reference (Original vector sometimes referred to as “O”). measurement of the response after adding the trial weight (Trial run vector known as “O+T”). inability to balance a rotor with this process will normally suggest that the correction weight is either not being added at the correct axial position or that the rotor will require correction in two or more planes.Graphical Balance Example .Single Plane Balancing Single plane balancing is the process of making balance corrections to a rotor in only one axial location. Figure 6 . Start Remove Trial Weight and Add Final Weight Measure Vibration & Determine if OK Measure Reference Vibration Determine Trial Weight magnitude and Location Determine Trim Weight if Required and Add Verify Acceptable Vibration Measure Response to Trial Weight and Determine Final Weight This will normally result in acceptable balance unless the machine was tremendously out of balance to start. then calculation of a final weight based on the shift of the vibration vector (“T”) and the amount of trial weight (TW). As an alternative to the graphical method. In cases where the data is accurate and linear. addition of a trial weight (TW). The process of single plane field balancing is shown below. the user is strongly advised to become proficient with a single plane graphical field balancing method prior to advancing to computer based solutions. or if there is a lot of nonlinear response or inconsistency in the data. The general way that single plane balancing is completed is by using a graphical method as shown in figure 6.
When the data is accurate and the system is linear.3.e. The methods described above. With this method. The graphical technique can be used. The use of multiple points will usually get more complicated than is easily resolved using the graphical method. vertical vs.… based on the number of measurement points used (no limit to the number of points) .2. With a single plane method using one sensor. so the weight change equals the trial weight.Computer based balancing routines will calculate solutions based on the input data. Balancing using a single measurement point should be capable of minimizing the vibration at that point. Since the different points will have different effects. However. The recommended method used to calculate balance correction weights (“shots”) when multiple sensors are used is the least squares numerical method. but it is common that different points (i. and is found by determining the amplitude change from O to O+T from a vector plot. it is necessary to use multiple measurement points and minimize the vibration at several points at the same time. are ideal when balance corrections are made to a simple rotor using a single vibration measurement location. and a single plane balance correction is desired. the response to balance corrections can be accurately calculated using the analytical method described in the following sections. horizontal) will have different reaction to a balance weight. However. or to use a numerical method. and the response with the trial weight is measured (O+T). use of a single measurement location will usually result in very low vibration at that location and higher vibration at another point. or unbalance in multiple planes. fluent use of graphical tools can provide huge benefit in reviewing balance responses particularly when a balance job has not gone well. Multiple Point Single Plane Balance The use of multiple measurement points now requires an analytical approach to best minimize the vibration at all points simultaneously. Before the numerical process can be described. The vibration change is called the trial response (T). and is the most common method used in balance programs. particularly the graphical method. When this occurs. This difference is generally the result of the unbalance correction not being added at the actual point of unbalance on the rotor. a trial weight is added (TW). the concept of vibration and weight changes must be well understood. This method uses a numerical procedure to reduce the sum of the squares of the amplitudes on each measurement point. Most practical field balancing will involve measurement at multiple locations on the machine. it will be necessary either to graphically review the response of each point. terms are defined for each of the values described above but by using subscripts to help accommodate conversion to a computer program: • • Vij = Vibration vector (amplitude and phase) measured at point i during run j Point i – Measurement points will be used with i = 1. the original vibration is measured (O). the highest amplitudes are more heavily “weighted” to help drive the general vibration severity down. The original vibration run will normally have no trial weight. When this is shifted over to a numerical process.
where a single complex variable contains both amplitude and phase data (actually real and imaginary or quadrature response). but this added complexity is required for using multiple points and/or multiple planes. 1-2 This must certainly sound like a seriously overcomplicated way to describe a simple balance calculation. Run 4.) Wj = Weight vector (amplitude and phase) installed when Run j was recorded o Note that W1 should normally be zero (no weights added on initial reference) o Weight addition can become a complicated accounting process with resulting confusion between trial/final weights added as well as trim weights. 0-2. Show a similar plot but with two weights (trial and final) and show the effect calculation for each run 0-1. followed by any additional sets of vibration data (Run 3. Using the graphical method. ΔVijk = The vibration vector change between Runs j and k at point i ΔWjk = The weight vector change between Runs j and k Using these definitions. Therefore. the resulting final weight as shown above would include both amplitude and phase definition. Please note that each value shown above is a vector quantity that includes amplitude and phase. The numerical method described tracks the phase positioning by using the combined amplitude and phase by using complex math variables. Recommendation is to document weights using clear notes regarding weights that have been added and removed throughout the process to prevent confusion. the balance calculations are as follows for use of a single vibration sensor and a single correction plane (refer to figure 6): Initial weight – W1=0 (no initial weight) Initial vibration – V11 = Original amplitude and phase (“O”) at point 1 for run 1 Trial weight – W2 = Trial weight magnitude and phase (“TW”) is the weight installed for run 2 Trial run vibration – V12 = Vibration with added weight (“O+T”) at point 1 for run 2 ΔW12 = W2 – W1 = W1 (“TW”) is the weight change from run 1 to run 2 ΔV112 = V12 – V11 (“T”) is the vibration change from run 1 to run 2 at point 1 Effect = ΔV112/ΔW12 = “T”/”TW” is the influence coefficient for point 1 Final Weight = FW (remove trial and add final weight) Calculated final weight = = FW Using this description. . The other item excluded from the description above is the determination of the phase angle for the final weight. is the balance coefficient and is the influence coefficient for sensor 1 due to the trial run in plane 1 (the only correction plane used). …. with the numerical method.• • • • Run j – Runs include the reference run (Run 1). the angle between T and O is used to determine the amount of shift required for the balance weight. as well as the trial run (Run 2).
60 @ 125 3.90 @ 215 0.70 @ 219 0. Table 1 .60 @ 103 Trial Run Vibration (added 170 gm @ 270°) 4.05 @ 99 4.7 @ 282° 312.1 @ 280° 559.5 @ 297° 586. The example below for a large motor balance demonstrates this well: Example – large motor with rotor unbalance. A single plane trial weight (static correction with identical weights added in phase on opposite ends) was used.55 @ 214 1. The final balance correction weights are calculated using the following matrix operation: . but rather a least squares optimization that will best minimize the general vibration for all sensors simultaneously. The same data was then used to calculate a combined correction weight for all the points. For sensors where low sensitivity exists.If the same process is repeated using multiple sensors using the graphical method or the calculation procedure described above. with the final weight calculation done using single point method for each sensor individually with the data shown in table 1. The variation in the calculated weights will be due to variations in the data accuracy or more likely due to the vibration at the sensor location not being the result of unbalance at the correction plane. the influence coefficient for each sensor can be determined by calculating the vector vibration change and the change in the trial weight as shown above with the results configured into an influence coefficient matrix as shown below: where The reference vibration vectors would also be arranged into a similar column matrix as: where Vi is the reference vibration amplitude/phase The final correction weight will be calculated relative to this reference vibration matrix. different final weight amounts and positions will be calculated for each sensor.Multiple Point Balance Calculation Data Sensor Description Outboard Horizontal Outboard Vertical Inboard Horizontal Inboard Vertical Combined weight Reference Vibration (no weights) 6.44 @ 113 Using all sensors-> Calculated Weight using inidividual points 585.5 @ 295° 570 @ 282° When multiple sensors are used. This combined weight is not the average of all the calculated individual weights. the calculated balance correction weight using that sensor may be very large (or small) and unrealistic.50 @ 219 0.
The initial trial run was used to calculate a correction weight of about 5 oz at 108°. which shows the nonlinearity in balance coefficients as the vibration amplitude is reduced. Once the process is converted to a computer program that can handle the math. The software is.Graphical Example of "Taking a New O" on Each Shot The example in figure 7 shows a classical example of balance coefficient change with vibration amplitude. This is done by using the vibration change and the weight change for each run. To reduce the errors and variations as much as possible. This is similar to the process of “taking a new O” as has been taught for years using the graphical method. if capable software is used that can handle the reference selection. and Run 1-Run 3. Figure 7 . or if for some reason you have one set of “bad” data for a given run. Actually if two shots are completed. A graphical description of this process is shown in figure 7. capable of dealing with calculation of these differences provided that each set of vibration data is documented with a corresponding set of weights. three sets of balance coefficients can be calculated including Run 1Run 2. However. Although . it is generally more accurate to base future calculations on the most recent runs since vibration levels are generally decreasing as weights are added.is the conjugate transpose of and is found by transposing the matrix and changing the sign on the imaginary component of each element. in concept. the software can be allowed to “take a new O” on every run through the process. This is accommodated since the balance coefficients are determined from vibration and weight changes. the equations will support calculating a new set of balance coefficients (actually influence coefficients) every time a weight is added to the shaft. The solution final weight M will produce the lowest combined vibration by minimizing the squares of each sensor value. Run 2-Run 3. This concept can provide some tremendous help particularly if there is a lot of non-linearity in the data.
but for some reason the outboard vertical appears not to be responsive to correction weight.Projected Vibration on Motor Balance Sensor Description Outboard Horizontal Outboard Vertical Inboard Horizontal Inboard Vertical Reference Vibration (no weights) 6. or else the vibration is not the result of unbalance. Using this approach. When that is done.05 @ 99 4. If the projected vibration is at or near the same amplitude as the reference vibration.18 @ 227 0. and the calculated correction weight as follows: Review of the projected vibration values will provide tremendous insight on the quality of the data and the linearity of the balance process. Once multiple sensors are used.55 @ 214 1. If the calculations include a single measurement and a single balance correction plane.this is properly calculated for the reference and trial run. the results of that balance shot were much less than desired. they will usually be minimized together as described above.60 @ 103 Projected Vibration (added 570 gm @ 282°) 0. If a “new O” would have been used numerically (use BC23 to calculate a trim correction from run 3). the last shot would have produced significantly better results. the reference vibration data. the calculated balance correction weight should produce zero vibration at that specific measurement point. Example of projected vibration – the previous example from table 1 is used to calculate the projected vibration after addition of the 570 gram weight at 282°: Table 2 .90 @ 215 0. some points will reduce better than others. Projecting Results One of the better methods for improving the balancing process is by using either manual or computer methods to predict or project the vibration that should result after a set of calculated weights are determined. It should be noted that the calculated balance correction weight using only that sensor also showed a . Review of the balance coefficients as the machine was getting smoother indicates that the initial balance coefficients used for the first two shots were off by about 30° compared to those determined from the final two runs.13 @ 192 The projected vibration results above show that the vibration on the rougher points should be well reduced. Continuing to add balance correction weights in an attempt to further reduce vibration once projected vibration amplitudes are not decreasing (along with the actual vibration after adding weights) is not advised. but there is no easy way to tell without projecting results from the balance data. there is inconsistency or non-linearity in the data.23 @ 163 0.94 @ 247 0. then the balance corrections are either not near the actual unbalance. Projected vibration results can be determined by using the influence coefficients. the projected vibration values should be approaching zero when using multiple measurement points.
Accuracy of the calculations will certainly be better with higher response provided that the vibration amplitude is not increased (possibly producing non-linearity).significantly different correction weight than the others. Other trial weight estimating methods include the API residual unbalance limit of ISO 1940. In contrast.5 (common for turbines and large motors) and G6. and G2. In all cases. the vibration on the machine is already high. but is always dependant on the weight of the rotating assembly and the speed of the shaft. Selecting a Trial Weight Selecting a trial weight can be a dangerous proposition. Trial weight selection can be done in several ways. if the phase lag is much different than expected. Adequate trial weight would be considered to be enough weight to produce a 10% change in vibration vector based on the original vibration. ISO 1940 limits are specified with a range of G limits with unbalance levels depending on weight and speed as follows: and Common G levels include the API limit (identical to G0. Jackson has suggested use of a trial weight that will produce a dynamic force that equals 10% of the static weight or less. As observed in figure 8. This would suggest that this point is not a good sensor location to use for the balance calculations. the 10% journal reaction method produces the highest trial weight amplitude for low shaft speed. This selection is not easy unless a reasonable amount of information is known about the machine. the trial weight above should be divided into two equal weights with one of the half weights added to each end of the rotor. and adding an excessive trial weight could increase the vibration amplitudes. the vibration won’t significantly increase. So it would be very nice to at least get in the correct quadrant for the trial weight location and add enough weight to get some response.2 mils change on a polar plot. In many cases. the trial response should be a minimum of 0. For instance if the original vibration is 2 mils. yet small enough that. lower speed and less . Some of these are overlaid in figure 8 for a 1000 Lb rotor. It then becomes conservative at higher shaft speeds. the ideal trial weight would be one that provides enough effect to calculate a final weight from. higher speed machinery will usually have lower ISO 1940 G levels specified.3 (typical for fans and pumps). In general. you normally don’t want to make a lot of trial runs without reducing the vibration. If the balance is to be a static correction (equal weight in phase on opposite ends of the rotor). If the vibration is already high.67). The ISO 1940 has a range of G values that correspond to different machine classes.
the trial weight should be located opposite the heavy spot (or weight removal at the heavy spot). Assuming a machine operates at least 10% below the 1st critical speed (2000 RPM in figure 9). or if you know if the machine operates above or below a critical. If a machine has a critical speed or structural natural frequency near the operating speed. Ideally.1G limit may be the most appropriate choice. It has been suggested by some that a “safe” . If there are oil film bearings or other damping elements. Figure 8 . The mechanical lag plot above will also suggests that phase lag angles will always be between 0 and 180°.Comparison of Trial Weight Calculation Methods Locating the First Trial Weight In addition to selecting the amount of trial weight to add. ALWAYS use the smallest trial weight necessary to get a useful effect and use a second trial run with more trial weight if it is necessary to increase the trial weight to get an effect. additional phase lag particularly on bearing housings can be present with an additional 10-60 degree lag. This lag is made up of the mechanical lag (shown in figure 9) plus any electrical lag. Unless previous balance experience is known for the machine. a trial weight size equivalent to the API limit (G0. The lag in figure 9 assumes that the unbalance is located at 0°. In general. Jackson’s recommendation of the 0.67) could be excessive. the mechanical phase lag should be less than 20° or so. a trial weight much larger than those suggested may not be adequate to get a good effect.precision equipment will often have much higher specified G levels. There should always be a fairly repeatable phase lag on a given machine at constant speed. So in the final conclusion. selecting the first trial weight location can be a guess. In other cases. the next critical item is the phase location of the trial weight.
the phase lag will shift back to near zero and increase to 90° at the 2nd critical and then towards 180° well above the 2nd critical at the balance plane that is dominant. Figure 9 . If several trial weights are required. trial and final weight. If the rotor speed is high enough that it approaches or passes a 2nd critical speed. It should be noted that in theory it is not reasonable to expect a phase lag to ever be higher than 180°. Here are some common problems: . This principal can be used for higher speed equipment to determine which end of the machine is likely the predominant contributor since the phase lag will generally return to the same value as the rotor approaches each critical speed. The most common problem is likely readings and trial weight locations confused on various runs. or if you are alternating between “final” and “trim” weight methods. If the process flows easily with a reference. there will usually be very good results.Mechanical Phase Lag Due to Critical Speed Difficulties with Single Plane Balancing Some balance jobs just don’t go well. it is very easy to make some bookkeeping errors regarding weight changes and what references to use.trial weight position when nothing is known about the machine is 90°! This phase lag will almost always produce a trial run that will at the worst produce similar amplitudes of vibration but appreciable phase change which is certainly adequate for balance calculations. or if there is a lot of nonlinearity (inconsistent balance coefficients).
thermally sensitive rotors. Unbalance not in correction planes – If the unbalance is at a different location than where corrections can be made. bent shafts. etc. or using an incorrect reference when using a trim calculation One of the most basic of problems is inaccurate data used in the balancing process. Thermal shifts – Caused by coupling alignment changes due to thermal growth. varying shaft speed. cocked bearings. the projected vibration will never decrease for all sensors Inaccurate data – Movements in the tachometer phase position or sensor locations. Improper documentation – Improperly documenting weight locations during the process. Variation in the amount of weight is generally pretty small since weights are either weighed or known. . etc.• • • • • • Balancing syndrome – Viewing every vibration problem as a balance problem and continuing to balance after the balance effort is not producing acceptable results Sources of vibration other than unbalance – Shaft misalignment. However. The influence of the amount and angular position of weights is shown below indicating that the angle for weights is far more important than the actual amount of the weight. etc. variation in the angular position of trial and final weights can be large particularly if you aren’t real careful when adding weights. lack of data averaging.
Non-Linearity of Response Anyone who has done much field balancing will know that balance coefficients will change if the vibration amplitudes start out very high. NDES – non-drive end seismic. The tachometer reference signal was located in phase with the vibration sensors. a rotor kit was used with a configuration as shown in figure 10.3 in/sec peak is likely to have non-linear response. This is the concept of “taking a new O” that has been taught for years in many balance courses. Once the rotor was extremely well balanced. Points were labeled as follows: DES – drive end seismic. As a general rule of thumb. . any machine with over 0. The rotor was balanced to extremely low levels in two correction planes (at the disks). DEX – drive end proximity. and NDEX – non-drive end proximity. The sample points documented included horizontal proximity probe and seismic (accelerometer integrated to displacement) at each bearing location. Some machines will be non-linear at far lower vibration amplitudes. To demonstrate this concept. What this means is that using the standard methods of balancing will result in continual overshoot/undershoot with successive weight calculation as the balance coefficients change unless a method is used to recalculate the balance coefficients as the machine gets smoother. unbalance weights were added to the drive end disk and vibration response was recorded.
In addition.Rotor Kit Layout Influence coefficients (mils/gram) were calculated for each amount of unbalance weight using the well balanced state as the reference with proximity probe readings runout compensated. the vibration response should increase in a straight line as the unbalance weight is increased.Figure 10 .0 as the vibration increases.5 to 3. the phase lag angles are shown to generally increase with increasing vibration. while the seismic sensor never showed over 2 mils pk-pk. It is easily observed from the plots that the balance coefficients are certainly nonlinear and strongly dependant on the vibration amplitude. with an additional phase lag of as much as 45°! With these sorts of variations. Figures 11 through 13 show the vibration response as well as the calculated influence coefficients separated into amplitude and phase plots. The influence coefficients show that the balance response can vary by a factor of 1. If the vibration response was linear. it becomes clear why a common rotor kit is not a good demonstrator for field balancing . The maximum vibration amplitude measured on the proximity probes was a little over 6 mils pk-pk for the highest applied unbalance.
This is the reason why “taking a new O” is commonly required for machines that start out very rough.unless it is very well balanced from the start and only small unbalance weights are used for demonstrations. However. Based on this example. it is possible to allow the software to accommodate this if appropriately programmed to handle the “new O” by conducting the balance calculations using any two balance runs where vibration and weight locations are known without having to start over. Figure 11 .Vibration Amplitude vs. as described above. a severely unbalance rotor may result in significant phase lag and large amplitude errors for calculated balance weights until the vibration is reduced to more linear ranges (see also figure 7). Unbalance Applied .
.Influence Coefficient Phase vs. The primary difference is that trial weights must be added to each correction plane to determine the balance response at each sensor to the various weight plane locations.Figure 12 .Influence Coefficient Magnitude vs. Trial Weight Size Multiplane Balancing Multiplane balancing is an extension of the same methods used for single plane. Trial Weight Size Figure 13 .
The simplest multiplane balancing can be accomplished using the static/couple method. For instance. normally on opposite ends of the machine. This method is required for non-symmetric or highly flexible rotors. rotor speeds. If two trial runs are used as required for a two plane balance. The trial weight changes are documented for each balance run and arranged in a square matrix which will be a 2x2 for a two plane balance as follows: where Wij = weight change (vector difference) for run i in plane j . the change in vibration matrix will be as follows including two columns each corresponding to a shot: The vibration change matrix includes the vector change between the 1st and 2nd runs (column 1) and the 2nd and 3rd runs (column 2). The balance calculations are done using the same math approach as before. a two plane balance problem can be separated into two single plane balances with good results. but expanded to accommodate the multiple plane data as shown below for a two plane correction with any number of vibration sensors: The vibration vectors would also be arranged into a column matrix as before: where Vi is the vibration amplitude/phase at sensor i for a given run The vibration change for all sensors and a single weight change is = - between two runs j and k. When the rotor is symmetric. Rows in this matrix correspond to individual sensors. Many more sensors are often used to help in the balance calculations and confidence along the way by reviewing the projected vibration for each balance correction prior to making a shot. Multiplane balancing in a more general sense involves simultaneously solving the balance problem by adding different correction weights in more than one correction plane and at different phase angles. This is a balancing process that can be used on reasonably symmetric rotors that requires the vibration readings on opposite ends of the rotor to be resolved into in phase (static) and out of phase (couple) readings. and vibration sensors. As a minimum. and can accommodate any number of correction planes. this matrix will be a 4x2 (4 rows x 2 columns). If a two plane balance is done with 4 sensors. there must be at least as many sensors as correction planes. a two plane correction will require at least two sensors.
the only difference between a “final” weight calculation and a “trim” run is whether the reference vibration used to calculate the correction weights is the actual original vibration (which will require removing trial weights added since the reference run) or the last measured vibration (which allows existing weights to stay in place and calculated weights are added to existing for a trim correction). This is a more general form of calculation for the influence coefficients that can be done for any combination of trial weight sets provided that the trial weight sets do not form a singular matrix. the following table shows examples of acceptable and unacceptable trial weight combinations for a two plane balance: Plane 1 Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 2 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 180° Trial Run 1 – 1 @ 0° Trial Run 2 – 2 @ 180° Trial Run 1 – 1 @ 0° Trial Run 2 – 0 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 0 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 0° Plane 2 Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 2 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 180° Trial Run 1 – 1 @ 0° Trial Run 2 – 2 @ 180° Trial Run 1 – 0 @ 0° Trial Run 2 – 1 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 0° Trial Run 1 – 1 @ 0° Trial Run 2 – 1 @ 180° Comment Not acceptable Same weights for both runs Not acceptable Trial run 2 is a linear multiple of trial run 1 Not acceptable Trial run 2 is identical to trial run 1 except for phase change Not acceptable Trial run 2 is a linear multiple of trial run 1 and same phase difference in both planes Acceptable Acceptable Acceptable Identical to a static move and couple move Once the and matrices are assembled. At this point. These weights will be the weights to be ADDED relative to the weights that were installed when was measured. A singular matrix would be one that will not invert due to the determinant being zero. The C matrix includes columns that correspond to correction planes and rows that correspond to sensors. the influence coefficients are calculated as ( found by transposing the term on the left below). . the final weights are calculated relative to any vibration reading set using the same approach as described for the single plane numerical method: is the correction weight set (plane 1 and plane 2) that will produce the lowest combined vibration by minimizing the squares of each sensor value used in the assembly.Once these matrices are assembled. and will result when the two rows in the matrix are linear multiples of each other. For example.
the static and couple corrections can be calculated and applied using two separate single plane calculations. If a final weight reduces the vibration significantly but not adequately. When this can be done. The vibration response can again be predicted using the reference vibration used in the calculation along with the influence coefficient matrix and the calculated correction weight column matrix. many power companies and equipment suppliers have large amounts of historical static/couple response data for broad families of turbines and generators. a smaller trim weight will be determined and the non-linearity of the response will not influence the trim run response as much. This occurs for several reasons that are partly due to historical practice and partly due to ease of balance weight positioning. then it would be preferred to leave the final weight in place and add a trim using the last run as the reference. To demonstrate the difference between using a two plane balance program using a static/couple weight addition and separate balance planes. power generation equipment (turbines and generators) have fairly symmetric rotors and can often be reviewed using static and couple resolution of vibration vectors. From a historical perspective. This concept really falls apart for non-symmetric rotors and for flexible rotors that operate above one or more critical speeds since the modal response of the rotor will not consistently follow expected static and couple characteristics. the non-linearity resulting from the original higher vibration will continue to influence the calculations.Errors in calculations and in response can always be minimized by using the lowest reference available. My personal preference is to always use the last two runs to recalculate new influences and using the lowest reference available to increase the accuracy. With current balance response calculation capabilities (balance programs) is makes sense to analyze unbalance in these types of rotors using two plane methods that allow simultaneous solving of both the static and couple response. For data input into a balancing program. the calculated weights would result in a static and couple shot opposed to separate weights in each correction plane. If another weight is calculated relative to the original reference. Static and Couple Method Adding trial weights to certain classes of machinery are usually done using combinations of static and couple (dynamic) corrections. the two trial runs below are identical and will produce the same calculation results: Static/Couple Method: Trial Run 1 – 100 gram @ 0° on drive end and 100 gram @ 0° on non-drive end (static move) . In addition. If the response is characterized using the original vibration and the final weight and weights calculated from the final weight shot. the typical plane 1 and plane 2 correction planes would be used to input the static and couple corrections and/or trial weights. When this is done.
Since the balance process is a bit more tricky with the multiplane approach. yet the software by design will automatically separate them out into influence coefficients at each sensor for each plane.Plane 1 = 100 @ 0° Trial Run 2 . Identical Weight Selection for 2 Plane Method: Trial Run 1 – 100 gram @ 0° on drive end and 100 gram @ 0° on non-drive end (static move) Trial Run 2 – 100 gram @ 0° on drive end and 100 gram @ 180° on non-drive end (couple) Program input – Plane 1 = Drive End Trial Run 1 . . except that the trial weight amount is ½ the amount calculated based on the entire rotor weight since ½ is added to each balance plane. and leave that weight in place if the vibration is lower followed by adding a single weight in one plane or the other to resolve the two plane response. If this is done. it will be very difficult to manually separate the response for hand calculation of the balance coefficients. When input and used in the balance program in this way.Plane 1 = 100 @ 0° Plane 2 = Non-drive End Plane 2 = 100 @ 0° Plane 2 = 100 @ 180° Calculated weights are added to the machine with Plane 1 added to the drive end correction plane and Plane 2 weight added to the non-drive end correction plane. Yet in cases where the vibration is excessive. it is generally much more logical (and significantly easier to review on vector plots) to add trial weights to each correction plane one at a time.Plane 1 = 0 Plane 2 = Couple Plane 2 = 0 Plane 2 = 100 @ 0° Calculated weights are added to the machine with Plane 1 added as static (same weight to both ends at same phase) and Plane 2 weight added as a couple (same weight to both ends with Nondrive end weight located 180° from the drive end couple weight). it may be wise to add a trial weight in one (or both) correction planes. This allows easy calculation of the balance coefficients for each plane on all sensors directly using the response data. either method will produce the same balance results.Trial Run 2 – 100 gram @ 0° on drive end and 100 gram @ 180° on non-drive end (couple) Program input – Plane 1 = Static Trial Run 1 . Selecting Multiplane Trial Weights and Phase Angles The same general rules will apply to selecting trial weights for multiplane problems.Plane 1 = 100 @ 0° Trial Run 2 .
For high speed rotors that operate above a critical speed. as the rotor approaches the 1st critical speed.The same methods described for single plane trial weights apply for two planes except that based on the total rotor weight the trial weight is divided by two so that only half of the rotor trial weight is located in a single plane. It is also common to use static/couple type shots as well depending on the initial reference vibration. the end of the rotor with the predominant unbalance will generally show a 180° phase shift back to near the low speed phase (well below the 1st critical) with a similar 90° shift as it approaches the 2nd critical. the phase change during a coastdown can be used to help identify the most probably location of unbalance. As the rotor approaches the 2nd critical speed (well above the 1st). The vibration was higher on the discharge end. In particular. The compressor operates below the 2nd critical speed.Inlet End of Compressor (Side away from unbalance) . Figure 14 . the phase lag between the heavy and high spots will increase from near zero to 90° at the critical speed then increase to near 180° well above. but the presence of high unbalance on the discharge end was confirmed by the phase response during a coastdown showing the phase below both the 1st and 2nd critical to be at approximately the same phase near 90°. Inspection during overhaul showed severe balance piston seal rubs and a loose impeller on the last (discharge end) impeller. This type of response is shown in figures 14 & 15 which were recorded on a centrifugal compressor that had a damaged impeller on the discharge end.
. Examples were presented to demonstrate the influence of non-linearity in the balance response as well as the influence of errors in weight placement. Theoretical description is also provided for calculating balance correction weights for single and multiple plane rotors. Application has been given for methods to eliminate the need for “taking a new O” which is a commonly taught process for dealing with balance non-linearity and data errors. Methods were described to reduce the impact of data variations during a balance process. The method described is the least squares method which is used to minimize a number of sensors simultaneously whether using single or multiplane balancing.Discharge End of Compressor (End with unbalance) Summary Methods have been described for completing field balancing using a combination of graphical and numerical methods. Emphasis on accuracy of weight addition angular position was made showing that a 5° angular position error is about equivalent to an 8% error in the amount of added weight.Figure 15 .
LI.. 1988.References 1. Rotordynamics of Turbomachinery. Balance Quality of Rotating Rigid Bodies. Ronald L.E. July 2007. Rotordynamics and Balancing. Tx. 535 Acorn Street. Fundamentals of Balancing. Eshleman. Jackson. Houston. 27th Edition. 3. Robert F. Willowbrook... Vibration Institute.E. Steidel. 1979. IL. Geneva Switzerland (ANSI S2. Jr. American Petroleum Institute. Vance. ....D.. CRC Standard Mathematical Tables. John Wiley & Sons. Schenck Trebel. Washington. New York. ISO 1940. John M. 1973. Ph. International Organization for Standardization. 2. 7. 8. Practical Vibration Primer. NY 11729. API Standard 617 “Special-purpose Centrifugal Compressors for Refinery Services”. CRC Press. DC.. 2nd Edition. P. Eshleman. 9.Ronald L.D. 5. An Introduction to Mechanical Vibrations. Gulf Publishing. Balancing of Rotating Machinery. Deer Park. IL. Charles. 2005. 4. 6. 1979. Willowbrook. P. John Wiley & Sons. New York. Vibration Institute. Ph.19).
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.